BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pyconde-pydata-2025//talk//UVPALT
BEGIN:VTIMEZONE
TZID:Europe/Berlin
BEGIN:DAYLIGHT
DTSTART:20240423T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20241027T030000
RDATE:20251026T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20250330T030000
RDATE:20260329T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:PDFs - When a thousand words are worth more than a picture (or tab
 le). - Caio Benatti Moretti
DTSTART;TZID=Europe/Berlin:20250423T151000
DTEND;TZID=Europe/Berlin:20250423T154000
DTSTAMP:20260818T075942Z
UID:pretalx-pyconde-pydata-2025-UVPALT@pretalx.com
DESCRIPTION:PDF\, a must-have in RAG systems\, ensures visual fidelity acr
 oss platforms and devices\, at the expense of compromising what would be t
 he core condition for computers to properly process and interpret text: se
 mantics. That means any logical arrangement of text\, upon rendering\, exp
 lodes into dummy visual shards of data that literally portrait the bigger 
 picture for the human eye to perceive\, but no longer convey the informati
 on computers should grasp. Such a bottleneck already makes proper ingestio
 n of text-only documents a big challenge\, let alone when tables or figure
 s come into play\, the ultimate nightmare for PDF parsers\, not to say dev
 elopers. The rest you must have already foreseen: a RAG system barfing unr
 eliable knowledge from bad chunks (based on regular PDF parsing)\, if thos
 e ever get to be retrieved from a vector database. In this talk you can ga
 ther some vision-driven insights on how to leverage the strengths of PDF a
 nd language models towards good chunks to be ingested. Or\, in other words
 \, how multimodal models can go beyond trivial reverse engineering by deco
 mposing tables into its building blocks\, in plain language\, as how those
  would be explained to another human\; or better yet\, as how humans would
  ask questions about such pieces of knowledge. And from such a strategy\, 
 we transfer the same rationale to figures. Come along\, gather some insigh
 ts\, and get inspired to break down tables and figures from your own PDFs\
 , and to improve retrieval in your RAG systems.
LOCATION:Hassium
URL:https://pretalx.com/pyconde-pydata-2025/talk/UVPALT/
END:VEVENT
END:VCALENDAR